探讨构建采取调强放疗(IMRT)＋后装治疗宫颈癌患者的预后列线图预测模型的方法，并建立该模型的预后危险分层系统。

选择2015年6月至2016年12月于江苏省肿瘤医院/江苏省肿瘤防治研究所/南京医科大学附属肿瘤医院进行IMRT＋后装治疗的258例宫颈癌患者为研究对象。采用Kaplan-Meier法，绘制不同临床因素宫颈癌患者总体生存(OS)曲线，并采用Log-rank检验进行比较。采用多因素Cox比例风险回归分析法，筛选影响采取IMRT＋后装治疗宫颈癌患者预后的独立危险因素，并采用R软件rms程序包，构建宫颈癌患者预后列线图预测模型，采用受试者工作特征(ROC)曲线及曲线下面积(AUC)与校准曲线，评价该列线图预测模型预测采取上述治疗策略宫颈癌患者的预后效能。采用递归分割分析(RPA)法，根据列线图模型，预测采取上述治疗策略宫颈癌患者预后风险评分，并构建宫颈癌患者预后危险分层系统。本研究遵循的程序符合江苏省肿瘤医院伦理委员会规定，通过该伦理委员会审查及批准(审批文号：2023科-快017)，并与所有患者签署临床研究知情同意书。

①对本组258例宫颈癌患者均采取IMRT＋后装治疗后，完全缓解(CR)率、部分缓解(PR)率及临床总有效率分别为74.4%(192/258)、20.9%(54/258)及95.3%(246/258)。②本组258例宫颈癌患者的1、3、5年OS率分别为93.8%、79.5%和64.0%。采取Log-rank检验进行宫颈癌患者预后影响因素的单因素分析结果显示，宫颈癌患者年龄(χ²＝4.25、P＝0.039)，病理学类型(χ²＝15.41、P<0.001)，国际妇产科联盟(FIGO)临床分期(χ²＝22.17、P<0.001)，是否伴淋巴结转移(χ²＝28.37、P<0.001)，同期化疗情况(χ²＝10.99、P＝0.004)、诱导＋序贯化疗情况(χ²＝14.34、P<0.001)及采取IMRT＋后装治疗近期疗效(χ²＝68.67、P<0.001)，均是影响其预后的因素。③多因素Cox比例风险回归分析结果显示，非鳞状上皮宫颈癌(HR＝2.404，95%CI：2.305～3.577，P＝0.005)，FIGO临床分期为Ⅲ、Ⅳ期(HR＝2.455、5.374，95%CI：2.386～3.609、3.221～6.507，P＝0.006、<0.001)，伴淋巴结转移(HR＝4.325，95%CI：2.189～6.420，P<0.001)，未进行同期化疗(HR＝1.730，95%CI：1.359～2.811，P＝0.040)，采取IMRT＋后装治疗后达PR及未缓解(HR＝1.779、3.227，95%CI：1.424～2.888、3.100～5.317，P＝0.047、<0.001)，均是影响采取IMRT＋后装治疗宫颈癌患者预后的独立危险因素。④根据上述多因素Cox比例风险回归分析结果构建预测采取IMRT＋后装治疗宫颈癌患者的预后列线图预测模型，预测采取IMRT＋后装治疗宫颈癌患者的3、5年OS率的ROC-AUC分别为0.827(95%CI：0.731～0.923，P<0.001)和0.789(95%CI：0.695～0.883，P<0.001)，该预测模型对采取IMRT＋后装治疗宫颈癌患者3、5年OS率的区分度较高。绘制该预测模型的校准曲线分析结果显示，其对采取上述治疗策略宫颈癌患者的3、5年OS率的校准曲线，均与理想曲线较接近，拟合反映良好，该预测模型预测采取IMRT＋后装治疗宫颈癌患者的3、5年OS率与实际OS率相对一致，校准度较好。⑤根据该列线图预测模型，预测采取IMRT＋后装治疗宫颈癌患者风险评分，采用RPA法构建宫颈癌患者预后危险分层系统，将其分为极低风险组(风险评分<138分)、低风险组(138分≤风险评分<214分)、中风险组(214分≤风险评分<274分)和高风险组(风险评分≥274分)。在不同FIGO临床分期宫颈癌患者中进行验证结果显示，该危险分层系统可区分FIGO临床分期为Ⅱ～Ⅳ期极低风险组、低风险组、中风险组与高风险组患者的3、5年OS率，差异均有统计学意义(P<0.05)。

本研究基于宫颈癌患者病理学类型、FIGO临床分期、是否伴淋巴结转移、同期化疗情况、采取IMRT＋后装治疗近期疗效这5项因素，构建预测采取IMRT＋后装治疗宫颈癌患者的预后列线图预测模型，该模型区分度、校准度均较好，可有效预测采取IMRT＋后装治疗宫颈癌患者预后。基于该预测模型所构建的宫颈癌患者预后危险分层系统亦具有一定临床价值。

To explore the method of constructing a prognostic nomogram prediction model for cervical cancer patients with intensity modulated radiation therapy (IMRT) and after-loading therapy, and establish prognostic risk stratification system based on this model.

A total of 258 cervical cancer patients who received IMRT and after-loading therapy in Jiangsu Cancer Hospital / Jiangsu Institute of Cancer Research / Affiliated Cancer Hospital of Nanjing Medical University from June 2015 to December 2016 were selected as study subjects. Kaplan-Meier method was used to plot the overall survival (OS) curves of cervical cancer patients with different clinical factors, and Log-rank test was used for comparison. Multiple Cox proportional risk regression analysis was used to screen independent risk factors that affect the prognosis of cervical cancer patients receiving IMRT and after-loading therapy, and rms package of R software was used to build nomogram prediction model, and receiver operating characteristic (ROC) curve and area under curve (AUC) and calibration curve were used to evaluate the predictive performance of the nomogram prediction model for predicting the prognosis of cervical cancer patients receiving IMRT and after-loading therapy. The recursive partition analysis (RPA) method was used to predict individual risk scores of cervical cancer patients receiving IMRT and after-loading therapy based on nomogram model, and prognostic risk stratification system of cervical cancer patients was constructed. The procedures followed in this study complied with the regulations of the Ethics Committee of Jiangsu Cancer Hospital and have been reviewed and approved by the Ethics Committee (Approval No. 2023-017). All patients signed clinical research consent form.

①Among 258 cervical cancer patients, complete remission (CR) rate after IMRT and after-loading therapy was 74.4% (192/258), partial remission (PR) rate was 20.9% (54/258), and total clinical effective rate was 95.3% (246/258). ②The 1-year, 3-year, and 5-year OS rates of 258 cervical cancer patients were 93.8%, 79.5%, and 64.0%, respectively. Univariate analysis of prognostic factors for cervical cancer patients by Log-rank test showed that age (χ²=4.25, P=0.039), pathological type (χ²=15.41, P<0.001), International Federation of Gynecology and Obstetrics (FIGO) stage (χ²=22.17, P<0.001), with or without lymph node metastasis (χ²=28.37, P<0.001), concurrent chemotherapy (χ²=10.99, P=0.004), induction and sequential chemotherapy (χ²=14.34, P<0.001), and short-term efficacy of IMRT and after-loading therapy (χ²=68.67, P<0.001) were all factors affecting the prognosis of cervical cancer patients receiving IMRT and after-loading therapy (χ²=68.672, P<0.001). ③Results of multivariate Cox proportional risk regression analysis showed that non squamous cell carcinoma cervical cancer patients (HR=2.404, 95%CI: 2.305-3.577, P=0.005), FIGO stages Ⅲ and Ⅳ (HR=2.455, 5.374; 95%CI: 2.386-3.609, 3.221-6.507; P=0.006, <0.001), with lymph node metastasis (HR=4.325, 95%CI: 2.189-6.420, P<0.001), lack of concurrent chemotherapy (HR=1.730, 95%CI: 1.359-2.811, P=0.040), PR and no response after IMRT and after-loading therapy were all independent risk factors affecting the prognosis of cervical cancer patients receiving IMRT and after-loading therapy. ④Nomogram prediction model was constructed based on results of the above multivariate Cox proportional risk regression analysis to predict the prognosis of cervical cancer patients receiving IMRT and after-loading therapy. ROC-AUC of the nomogram prediction model for predicting 3-year and 5-year OS rates of cervical cancer patients receiving IMRT and after-loading therapy were 0.827 (95%CI: 0.731-0.923, P<0.001) and 0.789 (95%CI: 0.695-0.883, P<0.001), respectively, indicating that the prediction model had high discriminability for 3-year and 5-year OS rates of cervical cancer patients receiving IMRT and after-loading therapy. The calibration curve analysis results of the predicted model showed that the calibration curves of 3-year and 5-year OS rates of cervical cancer patients receiving IMRT and after-loading therapy were close to the ideal curves, and the fitting was well reflected, indicating that the predicted 3-year OS rates of cervical cancer patients receiving IMRT and after-loading therapy were relatively consistent with the actual OS rates, and the calibration was good. ⑤Based on the risk score of cervical cancer patients predicted by the nomogram prediction model, prognostic risk stratification system for cervical cancer patients was constructed with the RPA method, in which all cervical cancer patients were divided into four groups: extremely low risk group (risk score<138 points), low risk group (138 points≤risk score<214 points), medium risk group (214 points≤risk score<274 points), and high risk group (risk score≥274 points). Validation results among cervical cancer patients with different FIGO stages showed that the risk stratification system can distinguish the 3-year and 5-year OS rates of the extremely low risk group, low risk group, medium risk group, and high risk group in FIGO stage Ⅱ-Ⅳ cervical cancer patients, and all differences were statistically significant (P<0.05).

This study constructs nomogram prediction model for predicting the prognosis of cervical cancer patients treated with IMRT and after-loading therapy based on five factors: pathological type, FIGO staging, with lymph node metastasis, concurrent chemotherapy, and short-term efficacy of IMRT and after-loading therapy, and the model has good discrimination and calibration, and can effectively predict the prognosis of cervical cancer patients after treatment of IMRT and after-loading therapy. The prognostic risk stratification system for cervical cancer patients constructed based on this prediction model has certain clinical value.